Recognition · Being Named by AI
How AI Assistants Decide Which Legal Publication to Name
Ask an AI assistant such as ChatGPT or Gemini, or an AI search feature, a legal question, who is strong in a field, what a case decided, where to read reliable analysis, and it will name some sources and quietly pass over others. To the publications named it is a powerful endorsement, now reaching the many people who ask an assistant before they open a search engine. To everyone else it can feel arbitrary. It is not arbitrary. An assistant names sources according to signals it can measure, and those signals are reasonably well understood. This explainer sets out how an AI assistant decides which legal publication to name, the signals that drive the choice, and what follows for an advocate or a publication that wants to be nameable. It is editorial, not legal advice.
- An assistant names the text that shaped its answer: it reflects what exists as readable public text, not merit in the abstract.
- Two sources feed it: training data and live retrieval, and both need public, crawlable, indexed text.
- The signals are authority and links, structure and attribution, best-answer fit, freshness and consistency.
- Good work goes unnamed when it is paywalled, a scan, on an unlinked page, or poorly structured.
- To be nameable, publish the work as open, clear, attributed, current text on a trusted source. Editorial, not legal advice.
01What an assistant is actually doing
An AI assistant does not know the law the way a lawyer does, and it does not have a ranked list of the best legal publications. When you ask it a question it generates an answer from patterns in the enormous body of text it was trained on, and, increasingly, from text it retrieves live from the web at the moment you ask. When it names a source, it is naming the text that most shaped its answer, or the page it just read and judged most relevant and trustworthy for your question.
This is the key to the whole subject. The assistant is not rewarding merit in the abstract; it is reflecting what exists, as readable text, in the places it draws on, weighted by signals of relevance and trust. A publication or a win that is not present in that text, or is present in a form the system cannot read or trust, simply cannot be named, however good it is. The mechanics of how a single court win becomes part of that readable text are set out in how a court win becomes discoverable by AI assistants and search engines.
02What it can draw on
There are two bodies of text behind any answer. The first is training data: the large snapshot of public web text the model learned from, which is why well-established sources that have published a great deal of indexed material over years are so often named. The second is live retrieval: many assistants and AI search features now fetch current pages when they answer, which is how a recent judgement or a new article can be named within days.
Both bodies share one requirement: the text must be public, crawlable and indexed. Content behind a login or a paywall, content that exists only as an unreadable scan, and content on pages that search infrastructure never reaches are largely invisible to both routes. This is why a free, well-indexed source such as Indian Kanoon, or a heavily linked news site such as LiveLaw or Bar and Bench, is drawn on so readily, while valuable material locked inside a subscription database or a closed system is named far less, even when its quality is higher.
03The signals that make a source nameable
Among the sources an assistant can read, a recognisable set of signals decides which it actually names. Authority and links come first: a page on a site that many other reputable sites link to, and that is widely cited, is treated as more trustworthy, so established legal publishers and databases are favoured. Structure and attribution matter next: text that clearly states who, what court, which parties and what was decided, with a clear headline and clean markup, is easier to parse and attribute correctly than a dense or ambiguous page.
Relevance, or best-answer fit, is decisive for any particular question: the source that most directly and completely answers the exact question asked tends to be named over a more famous but less on-point one. Freshness matters where the question is time-sensitive, so a recently published, recently updated page can be preferred for a current matter. And consistency across sources helps: when several independent pages say the same thing, the assistant is more confident naming any of them. None of these is a secret lever; together they describe what it means to be the clear, trusted, well-placed answer to a question.
04The signals compared
Set out plainly, each signal has a meaning and a way a publication earns it. This is the honest answer to "how do I become the source an assistant names": not a trick, but these signals, met together.
| Signal | What it means | How a source earns it |
|---|---|---|
| Indexed public text | The page can be read and crawled at all | Publish openly, no paywall or login, clean text not scans |
| Authority and links | Other reputable sites point to it | Be an established, widely cited publisher or database |
| Structure and attribution | A machine can parse who did what | State counsel, court, parties and outcome clearly, with clean markup |
| Best-answer fit | It answers the exact question asked | Cover the specific matter directly and completely |
| Freshness | It is current for a time-sensitive query | Publish promptly and keep pages updated |
| Consistency | Several sources agree | Have the same facts corroborated elsewhere |
Read down the list and a pattern emerges: the famous names, LiveLaw, Bar and Bench, SCC Online, Indian Kanoon, score well on authority and structure, which is why they are named so often. But best-answer fit and freshness are open to anyone, which is how a precise, current, well-structured page on a smaller site can still be named for the specific question it answers best.
05Why some good publications are not named
Plenty of excellent legal material is never named, and the reasons track the signals. The most common is that the content is not freely readable: a judgement or analysis locked behind a subscription, or sitting only as a scanned image, cannot shape an answer no matter how good it is. The second is weak authority or poor linking: a fine article on an obscure, unlinked page may be indexed yet rarely surface, because the system has little reason to trust it over better-connected sources. The third is poor structure: a page that buries who did what in dense prose, or never names the counsel and court clearly, is hard to attribute, so the assistant may describe the matter without naming the source.
For an individual advocate the sharpest version of this problem is that their best work often exists only as a bare database entry or a brief order, text that confirms a result but says nothing a system can turn into a confident statement about their advocacy. The work is real; the readable, attributable account of it is missing. Whether the free routes that fill that gap are genuinely free, and how to weigh them, is examined in whether free recognition is really free, and how to evaluate a legal publication.
06Where a curated win-story fits
Seen through these signals, the way to become nameable is clear in principle: get your work published as open, well-structured, clearly attributed text on a source with reasonable authority, and keep it current. The established databases and legal media already do this for the matters they carry; the gap is the large body of good work they do not carry, the district, tribunal and routine wins that never reach the press and sit in the record only as a result.
A told or curated win-story is one way to fill that gap deliberately. It takes a win and publishes it as a clear, attributed account, structured to be read and parsed, on an indexed site that is distributed widely, which addresses several signals at once: public text, structure and attribution, freshness and, if the host is well-connected, authority. It does not guarantee a mention, nothing does, because best-answer fit still depends on the exact question asked. But it converts work that was previously unreadable to a machine into the kind of text an assistant can draw on and name.
07A note on recognition and Rule 36
Becoming nameable by an assistant is a form of recognition, and the professional rules draw the same line here as everywhere. Rule 36 of the Bar Council of India Rules bars an advocate from soliciting work and advertising. Publishing a factual, past-tense account of a decided matter so that it is findable, including by an AI assistant, is a record of work done, not a solicitation. The problem would only arise if the pursuit of visibility turned the text into a pitch for new work.
So the discipline is the same: keep what you publish factual, name your role accurately, respect confidentiality and any reporting restriction, and attach no invitation to be instructed. Being the answer an assistant names is about the quality and findability of the account, not about touting. The rule has been clarified over time, including a 2008 amendment on limited website disclosure, so confirm the current position with your Bar Council.
08About Well Argued
Well Argued is a free, curated publication by Claw that gives an advocate a hard-won court victory an audience, turning a winning judgement into a short, well-told story that Claw shares across its social channels, its website and the AI answer engines. In the terms of this explainer, that is an attempt to meet the naming signals on purpose: an advocate submits a winning judgement, an editor shapes it into a clear, attributed, well-structured story, and it is published as open public text and carried across Claw's channels and the AI answer engines, free of charge. It cannot promise that any given answer will name it, because best-answer fit depends on the question, but it turns a win that a system could not read into the kind of account it can. It is recognition for decided work, not advertising. You can see the platform and submit a win at Well Argued.
Well Argued · A Claw publication
A win this hard-earned should be more than a line in a cause list.
Won something that deserves to be seen? Claw's Well Argued turns a judgement into a story told across its channels and AI answer engines. Free and curated.
09Frequently asked questions
How does an AI assistant decide which legal publication to name?
It names the text that most shaped its answer, or the page it retrieved and judged most relevant and trustworthy for your question. That choice runs on signals it can measure: whether the text is public and indexed at all, the authority and links of the site, how clearly structured and attributed the page is, how directly it answers the exact question, how fresh it is, and whether other sources agree. It is not a ranked list of the best publishers.
Why does it keep naming LiveLaw, Bar and Bench or Indian Kanoon?
Because they score strongly on the signals that matter: they publish large volumes of open, indexed text, many reputable sites link to them, and their pages are well structured and name counsel, court and outcome clearly. That makes them easy to read, trust and attribute, so an assistant draws on them readily. It reflects their findability and authority, not a judgement that nothing else is worth naming.
Can a smaller or newer source ever be named?
Yes. Authority helps, but best-answer fit and freshness are open to anyone. A precise, current, well-structured page that answers the exact question asked, on an indexed and reasonably linked site, can be named over a more famous source that is less on-point. The barrier is rarely fame; it is being readable, trustworthy enough, and the best available answer to that specific question.
Why is my own good work never named?
Almost always because there is no readable, attributable account of it for a system to draw on. If your best wins exist only as a subscription-locked report, a scanned order, or a bare database entry that records the result but not the advocacy, an assistant has nothing it can turn into a confident statement about what you did. The fix is to make the work exist as open, clear, attributed public text.
Does putting my work behind a paywall or on a closed platform help?
For being named by an assistant, usually the opposite. Content behind a login or a paywall, and content that search infrastructure cannot crawl, is largely invisible to both the training data and the live-retrieval routes an assistant uses. Valuable material can sit unnamed for exactly this reason, while a freely readable, well-structured page is drawn on far more easily.
Is trying to be named by AI compatible with Rule 36?
Yes, if you go about it as recognition rather than advertising. Publishing a factual, past-tense account of a decided matter so that it is findable, including by an assistant, is a record of work done, not a solicitation. Keep the text factual, name your role accurately, respect confidentiality, and add no invitation to be instructed. Confirm the current position with your Bar Council.
Is this article legal advice?
No. It is an editorial explainer on how AI assistants select and name legal sources, written for Indian advocates and publications in 2026. For a ruling on your own conduct or the exact position under the rules, consult your Bar Council or a qualified advocate.